Santayana, yadda, yadda… This afternoon I’m going to write about Andrew Hamilton and Quentin Wheeler’s “Taxonomy and Why History of Science Matters for Science: A Case Study”, which derives lessons from the history of numerical taxonomy (phenetics) for the future of DNA bar-coding. I wasn’t aware of phenetics, which seems to have been a mid-century attempt to measure living things and then group them without recourse to any overarching theory. This has intriguing parallels to the mathematics of the Bourbaki collective that I won’t go into (and don’t actually know much about), but I just wanted to throw that out there. The big point here is that the phenetics movement precipitously collapsed after its haphazard data-collecting failed to produce a believable taxonomy, and the authors argue that the same could happen to DNA bar-coding, which uses DNA arrangements to draw relationships between different organisms.
The first point I’d like to address is the use of the lesson from history. In my first post in this series, I discussed the use of history by filmmakers. Here I’m more reminded of the constant use of history in politics, which is notoriously dicey in its deployment of analogies with past events. Here in America we’re being subjected to fairly sophisticated historical analyses on a daily basis as the Presidential campaign goes forward. Inevitably, we learn why the strategies being deployed are similar to Reagan vs. Carter in 1980 or Nixon vs. Kennedy in 1960, and so on… I think, as with filmmakers using history, this is both inevitable and healthy, but there’s a difference. Here we’re not dealing with a flow of ideas and methods over time, but with isolated lessons. It’s like studying past Super Bowls or games of chess to provide a palette or menu of possible strategies.
So, I guess this is fair, but what’s really going on is the use of history as an illustration of concept. Hamilton and Wheeler are a little apologetic that this isn’t really history in a proper sense. The use of historical examples to make arguments about what might or might not constitute valuable work in the present is a tad Whiggish, I suppose, but I have no real qualms about it, since the phenetics example is from a similar enough scientific culture to the present to make the analogy worthwhile. What strikes me is the redundancy of the whole exercise. Isn’t it valid from a philosophical or conceptual standpoint just to say, “your data is meaningless because it exists in an intellectual vaccum”? Saying “you’re a lot like those phenetics guys, and we all know what happened to them” just seems like belaboring the point.
There are actually more interesting, and possibly more fundamental historical analogies to draw upon here outside the realm of biology. There’s a classic debate in economics surrounding a 1947 paper by Tjalling Koopmans called “Measurement without Theory” that could just as easily be used (and which contains a little historical lesson-drawing of its own!).* Actually, that’s even a more interesting case, because the theory in question was some pretty abstract stuff. On the other hand, Francis Bacon advocated gathering all kinds of specimens and histories and then playing around with them to see what principles emerged. And that’s valid, too.
In the case of DNA bar coding, there are extenuating factors involved that Hamilton and Wheeler are glad to present, but don’t really take into account. First is the question of robustness. They point out that DNA bar coding may be useful in cases where species such as bacteria are morphologically ambiguous, but question whether it should be used to make taxonomical statements in areas where the terrain is better known. This, even though the bar coding proponents openly say that their claims are provisional and subject to correspondence with other forms of evidence that are more firmly grounded in evolutionary theory. (The article left the impression that bar coding advocates suggested that bar coding should be a definitive or sole method, but I’d like to see more before seeing the bar coders as advocating that point of view).
The main sticking point seems to be that there are a lot of species out there that aren’t taxonomized (right word?), and that DNA bar coding could aid in the effort to do so. So the question is whether it’s better to have an automatically-generated but theoretically-unstable taxonomy or none at all. So long as the means of the organisms’ classification remains known, and the method is thought to be fairly reliable (which seems possible or even likely), I don’t see the problem, so long as the value of the exercise is deemed worthwhile measured against the expected reliability of the technique and the expense involved.
Hamilton and Wheeler, though, I think are less concerned that the cost-benefit analysis here is inadequate (which is what they claim), and more that the status of the taxonomy created by DNA bar coding will be lost and taken as fact or as giving us information the procedure is not designed to provide, which gets into arguments about scientism. I’ve always thought concerns about scientism in the history of science have been way overcooked, because the “asserted fact” often seems to be equated with “existence of published text” even though there are loads of sociological means of signaling the status of claims as speculative, likely, certain, etc…, by which, crucially, the bar coders seem committed to abiding. All in all, I call foul on this history lesson. Reading the evidence they themselves present, the authors have not adequately convinced me that we are doomed to repeat history.
*Philip Mirowski has epistemologically linked Bourbaki with Koopmans in his book Machine Dreams, so all this is actually sort of tied together. I’m not going to go into this here (it’s incredibly complex); suffice it to say Mirowski loathes both Koopmans and Bourbaki.